Spatialization Modeling Of Population Based On Globeland30 Land Cover

Population growth is one of the main driving forces for increasing the risk of resource and environment,natural disasters and climate change,so obtaining the information of population spatial distribution quickly and accurately is beneficial to solving the problem of natural resources and environment.The statistical population data,which takes administrative division as the basic unit,reflects the general characteristics of the cell.The degree of data refinement is low.The spatial and temporal resolution is low.The inconsistency between the geographic units causes it to be difficult to integrate with other natural data.The spatialization of population data can realize the conversion of administrative unit population data,which can help to broaden the application depth and breadth of demographic data.Land use data is regarded as a complex that covers much information about the spatial distribution of population,so that spatialization of population data based on land use/land cover is the most commonly used method.However,the accuracy of the method depends on the type and quality of the land cover data,spatialization methods,and modeling units,and it is impossible to distinguish the population distribution differences within the unit.On this basis,based on the land cover data of GlobeLand30 in 2010,the method of multiple linear regression and Geographical Weighted Regression(GWR)are used to spatialization of population data based on different types of land cover on the city and county level,and analyzing the results.At the same time,it proposes a method of spatialization of population data based on population potential to represent the internal difference of population.Firstly,the modeling factors are determined by the correlation analysis between the area index of the cover type and the population density.On the county level and the municipal level,the spatial model of single factor and multi factor population data are constructed based on the multiple linear regression method,and the error analysis is carried out.The experimental results show that the method of population data spatialization based on multiple linear regression is not suitable for large-scale fitting.The main reason is that the whole research area is regarded as a "homogeneous" unit,and the heterogeneity of the region is ignored,and the county level error is higher than the municipal level,indicating that the spatial heterogeneity of the county level is stronger than that of the municipal level.Then,using the same experimental scheme,the spatial model of population data is constructed based on the GWR method,and compared with the multiple linear regression model to verify the fitting effect of GWR.The experimental results show that the model accuracy of the GWR method is higher than the multiple linear regression method,which can overcome the spatial heterogeneity of the region.At the same time,the multi factor modeling is higher than the single factor accuracy,which shows that the increase of the modeling factor can effectively improve the model accuracy.The accuracy of the spatial data of the population data based on artificial surface and cultivated land is tested at county level units,and the results show that the model is reliable.Finally,the population potential surface is introduced to reflect the degree of population agglomeration on the artificial surface of the research unit,and the corresponding population surface is obtained by setting different radii.The model is constructed using the GWR model,and the error analysis and test are carried out.The experimental results show that the radiating radius is 20 km,the modeling effect is the best,the fitting accuracy is R2 =0.901,and the average relative error of population is 27.9%.The error of the county level test is 41.3% which is significantly higher than that of the municipal level,indicating that the population potential of 20 km can reflect the population concentration of the city level,but it does not meet the county-level situation,indicating that the population potential has a scale effect.